How AI can reveal political language evolution, patterns, and biases from 123 years of federal budget speeches in Australia

Visual Stories Team | The Age
1 June 2023

Intro to NLP

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Risus at ultrices mi tempus imperdiet. Dignissim diam quis enim lobortis scelerisque fermentum dui faucibus. Tempor nec feugiat nisl pretium fusce id velit. Ut lectus arcu bibendum at varius vel pharetra vel. Rhoncus urna neque viverra justo nec ultrices dui sapien eget. Vestibulum sed arcu non odio euismod lacinia at quis risus.

Toy case study

Toy case study of how Natural Language Processing (NLP) analyses text.

Showcase of how modern computing and machine learning can now easily analyse millions of words and historical patterns ie. use a simple case study of language processing to introduce NLPs then dive into budget speech analysis

How NLP text analysis works

Step through of how NLP models analyse text using tokenising to parse individual words, verbage, language, and keywords, such as identifying nouns, pronouns, verbs, and keywords (Universal Parts of Speech, UPOS).

Token/tokenising = individual words
UPOS = Universal Parts of Speech (word classification)
Lemma/lemmatising = root word form of each token
Tagging
Dependency Parsing Annotation

Topic selection

Select three global and political topics that most interest you

Data dive into topics reader chooses (below outputs)

Defence

Climate

Housing

Immigration

Environment

Jobs (unemployment)

Tax (breaks or increases)

Pension

Aged care

Transport

Education

Trade

Budget - Debt / surplus

Technology

Manufacturing

Health

Hospitals

Crime

Espionage

Timeline analysis

Timeline analysis of the last 123 years of budget speeches using chosen topics

Highlight key global events in timeline to show how different parties and Aus politics in general responded to global events eg Greta Thunberg’s birth
how language shifts over time and between parties
how parties favour keywords or language surrounding key topics
which parties react to global events/issues

Sentiment analysis of language

Sentiment analysis of tone

Analysis of positive vs negative discourse - how language changes over time, which parties are more -/+ on issues, show how certain topics are systematically linked to negative language eg. illegal immigration, unskilled immigrants, etc.

Defence

Tax

Deep dive example

Lessons from Natural Language Processing

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Risus at ultrices mi tempus imperdiet. Dignissim diam quis enim lobortis scelerisque fermentum dui faucibus. Tempor nec feugiat nisl pretium fusce id velit. Ut lectus arcu bibendum at varius vel pharetra vel. Rhoncus urna neque viverra justo nec ultrices dui sapien eget. Vestibulum sed arcu non odio euismod lacinia at quis risus.

Nunc mi ipsum faucibus vitae aliquet nec ullamcorper. Tempus urna et pharetra pharetra. Diam quam nulla porttitor massa id neque aliquam vestibulum morbi. Blandit cursus risus at ultrices mi tempus imperdiet. Mauris ultrices eros in cursus turpis massa tincidunt dui. Cras fermentum odio eu feugiat. Vitae tortor condimentum lacinia quis vel eros donec. Scelerisque mauris pellentesque pulvinar pellentesque habitant morbi tristique senectus. Tempor nec feugiat nisl pretium. Vitae tempus quam pellentesque nec nam aliquam.

Credits

DEVELOPER/3D VISUALISATION Richard Lama, Nathanael Scott, Daniel Carter PHOTOS Francis Kokoroko DATA Matt Malishev DESIGN Mark Stehle, Luis Araujo, Kathleen Adele VIDEO Francis Kokoroko PRODUCTION Matthew Absalom-Wong EDITORS Nick Ralston, Lia Timson